Accurate Compression of Tabulated Chemistry Models with Partition of Unity Networks

Armstrong, Elizabeth; Hansen, Michael A.; Knaus, Robert C.; Trask, Nathaniel A.; Hewson, John C.; Sutherland, James C.

doi:10.1080/00102202.2022.2102908

Accurate Compression of Tabulated Chemistry Models with Partition of Unity Networks

Journal Article · Sun Aug 07 00:00:00 EDT 2022 · Combustion Science and Technology

DOI:https://doi.org/10.1080/00102202.2022.2102908· OSTI ID:1883195

^[1]; Hansen, Michael A. ^[2]; Knaus, Robert C. ^[2]; Trask, Nathaniel A. ^[2]; Hewson, John C. ^[2]; Sutherland, James C. ^[1]

University of Utah, Salt Lake City, UT (United States)
Sandia National Laboratories (SNL-NM), Albuquerque, NM (United States)

Tabulated chemistry models are widely used to simulate large-scale turbulent fires in applications including energy generation and fire safety. Tabulation via piecewise Cartesian interpolation suffers from the curse-of-dimensionality, leading to a prohibitive exponential growth in parameters and memory usage as more dimensions are considered. Artificial neural networks (ANNs) have attracted attention for constructing surrogates for chemistry models due to their ability to perform high-dimensional approximation. However, due to well-known pathologies regarding the realization of suboptimal local minima during training, in practice they do not converge and provide unreliable accuracy. Partition of unity networks (POUnets) are a recently introduced family of ANNs which preserve notions of convergence while performing high-dimensional approximation, discovering a mesh-free partition of space which may be used to perform optimal polynomial approximation. In this work, we assess their performance with respect to accuracy and model complexity in reconstructing unstructured flamelet data representative of nonadiabatic pool fire models. Our results show that POUnets can provide the desirable accuracy of classical spline-based interpolants with the low memory footprint of traditional ANNs while converging faster to significantly lower errors than ANNs. For example, we observe POUnets obtaining target accuracies in two dimensions with 40 to 50 times less memory and roughly double the compression in three dimensions. We also address the practical matter of efficiently training accurate POUnets by studying convergence over key hyperparameters, the impact of partition/basis formulation, and the sensitivity to initialization.

View Accepted Manuscript (DOE)

Research Organization:: Sandia National Laboratories (SNL-NM), Albuquerque, NM (United States)

Sponsoring Organization:: National Science Foundation (NSF); USDOE National Nuclear Security Administration (NNSA)

Grant/Contract Number:: NA0003525

OSTI ID:: 1883195

Alternate ID(s):: OSTI ID: 1908749

Report Number(s):: SAND2022-9641J; 708276

Journal Information:: Combustion Science and Technology, Journal Name: Combustion Science and Technology Journal Issue: 6 Vol. 196; ISSN 0010-2202

Publisher:: Taylor & FrancisCopyright Statement

Country of Publication:: United States

Language:: English

References (20)

Artificial neural network implementation of chemistry with pdf simulation of H2/CO2 flames Christo, F. Combustion and Flame, Vol. 106, Issue 4 https://doi.org/10.1016/0010-2180(95)00250-2	journal	September 1996
A single-step time-integrator of a methane–air chemical system using artificial neural networks Blasco, J. A.; Fueyo, N.; Larroya, J. C. Computers & Chemical Engineering, Vol. 23, Issue 9 https://doi.org/10.1016/S0098-1354(99)00278-1	journal	November 1999
Laminar flamelet modeling of a turbulent CH4/H2/N2 jet diffusion flame using artificial neural networks Emami, M. D.; Eshghinejad Fard, A. Applied Mathematical Modelling, Vol. 36, Issue 5 https://doi.org/10.1016/j.apm.2011.08.012	journal	May 2012
Chemical explosive mode analysis for a turbulent lifted ethylene jet flame in highly-heated coflow Luo, Zhaoyu; Yoo, Chun Sang; Richardson, Edward S. Combustion and Flame, Vol. 159, Issue 1 https://doi.org/10.1016/j.combustflame.2011.05.023	journal	January 2012
Tabulation of combustion chemistry via Artificial Neural Networks (ANNs): Methodology and application to LES-PDF simulation of Sydney flame L Franke, Lucas L. C.; Chatzopoulos, Athanasios K.; Rigopoulos, Stelios Combustion and Flame, Vol. 185 https://doi.org/10.1016/j.combustflame.2017.07.014	journal	November 2017
Experimental and numerical study of water sprayed turbulent combustion: Proposal of a neural network modeling for five-dimensional flamelet approach Honzawa, Takafumi; Kai, Reo; Hori, Kotaro Energy and AI, Vol. 5 https://doi.org/10.1016/j.egyai.2021.100076	journal	September 2021
Investigation of lengthscales, scalar dissipation, and flame orientation in a piloted diffusion flame by LES Kempf, A.; Flemming, F.; Janicka, J. Proceedings of the Combustion Institute, Vol. 30, Issue 1 https://doi.org/10.1016/j.proci.2004.08.182	journal	January 2005
Optimal artificial neural networks and tabulation methods for chemistry representation in LES of a bluff-body swirl-stabilized flame Ihme, Matthias; Schmitt, Christoph; Pitsch, Heinz Proceedings of the Combustion Institute, Vol. 32, Issue 1 https://doi.org/10.1016/j.proci.2008.06.100	journal	January 2009
A chemistry tabulation approach via Rate-Controlled Constrained Equilibrium (RCCE) and Artificial Neural Networks (ANNs), with application to turbulent non-premixed CH4/H2/N2 flames Chatzopoulos, A. K.; Rigopoulos, Stelios Proceedings of the Combustion Institute, Vol. 34, Issue 1 https://doi.org/10.1016/j.proci.2012.06.057	journal	January 2013
Small scales, many species and the manifold challenges of turbulent combustion Pope, Stephen B. Proceedings of the Combustion Institute, Vol. 34, Issue 1 https://doi.org/10.1016/j.proci.2012.09.009	journal	January 2013
Efficient bifurcation and tabulation of multi-dimensional combustion manifolds using deep mixture of experts: An a priori study Owoyele, Opeoluwa; Kundu, Prithwish; Pal, Pinaki Proceedings of the Combustion Institute, Vol. 38, Issue 4 https://doi.org/10.1016/j.proci.2020.09.006	journal	January 2021
Turbulent Combustion Peters, Norbert Cambridge University Press https://doi.org/10.1017/CBO9780511612701	book	January 2010
Predicting large-scale pool fire dynamics using an unsteady flamelet- and large-eddy simulation-based model suite Domino, Stefan P.; Hewson, John; Knaus, Robert Physics of Fluids, Vol. 33, Issue 8 https://doi.org/10.1063/5.0060267	journal	August 2021
Characterizing Tradeoffs in Memory, Accuracy, and Speed for Chemistry Tabulation Techniques Armstrong, Elizabeth; Hewson, John C.; Sutherland, James C. Combustion Science and Technology https://doi.org/10.1080/00102202.2022.2028780	journal	January 2022
Adaptive chemistry lookup tables for combustion simulations using optimal B-spline interpolants Bode, Mathis; Collier, Nathan; Bisetti, Fabrizio Combustion Theory and Modelling, Vol. 23, Issue 4 https://doi.org/10.1080/13647830.2019.1583379	journal	February 2019
A self-organizing-map approach to chemistry representation in combustion applications Blasco, J.; Fueyo, N.; Dopazo, C. Combustion Theory and Modelling, Vol. 4, Issue 1 https://doi.org/10.1088/1364-7830/4/1/304	journal	March 2000
The Gap between Theory and Practice in Function Approximation with Deep Neural Networks Adcock, Ben; Dexter, Nick SIAM Journal on Mathematics of Data Science, Vol. 3, Issue 2 https://doi.org/10.1137/20M131309X	journal	January 2021
Deep ReLU networks and high-order finite element methods Opschoor, Joost A. A.; Petersen, Philipp C.; Schwab, Christoph Analysis and Applications, Vol. 18, Issue 05 https://doi.org/10.1142/S0219530519410136	journal	February 2020
Application of deep artificial neural networks to multi-dimensional flamelet libraries and spray flames Owoyele, Opeoluwa; Kundu, Prithwish; Ameen, Muhsin M. International Journal of Engine Research, Vol. 21, Issue 1 https://doi.org/10.1177/1468087419837770	journal	March 2019
Acceleration of Chemical Kinetics Computation with the Learned Intelligent Tabulation (LIT) Method Haghshenas, Majid; Mitra, Peetak; Santo, Niccolò Dal Energies, Vol. 14, Issue 23 https://doi.org/10.3390/en14237851	journal	November 2021

Similar Records

Characterizing Tradeoffs in Memory, Accuracy, and Speed for Chemistry Tabulation Techniques

Journal Article · Tue Jan 25 19:00:00 EST 2022 · Combustion Science and Technology · OSTI ID:1882871

An a priori evaluation of a principal component and artificial neural network based combustion model in diesel engine conditions

Journal Article · Fri Aug 28 20:00:00 EDT 2020 · Proceedings of the Combustion Institute · OSTI ID:1845097

Application of deep artificial neural networks to multi-dimensional flamelet libraries and spray flames

Journal Article · Thu Mar 28 20:00:00 EDT 2019 · International Journal of Engine Research · OSTI ID:1605138

Related Subjects

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CHEMISTRY
artificial neural network
chemistry tabulation
partition of unity

Accurate Compression of Tabulated Chemistry Models with Partition of Unity Networks

Citation Formats

References (20)

Similar Records

Related Subjects